|No such thing as a small change|
Re^3: RFC: I rewrote a custom sort function, but not sure if it is the best way.by roboticus (Canon)
|on Mar 03, 2013 at 15:46 UTC||Need Help??|
Regarding "my_sort_index": Oh, yeah, I totally forgot to write that.
Well, when writing comparison routines, you first want to determine what all the cases are that you want to handle, and what to do for each case. For ensuring that strings beginning with "index." sort to the top of the list, there are four cases:
So, here's my (untested!) version of the my_sort_index routine:
We could enforce the order in the fourth case, but it would make the subroutine a bit less useful. Why? It makes the subroutine harder to use with other subroutines for a sorting job, because it forces an order beyond just making sure that strings prefixed with "index." go first. In other words, if neither string began with index, it would use a different ordering than if both strings began with index. That can be problem.
Suppose you want all strings with "index." to sort to the top of the list, but you want everything sorted in reverse order beyond that, you'd have to have another version of your code. With the code we have now, though, we could do:
Regarding the Schwartzian transform: I wanted you to be aware of the technique, because it can let you do some pretty interesting things. When I use them, it's generally because I think to myself, "Oh, this would be *so* much easier to sort if everything looked like XYZ". Then I try to figure out an easy way to take what I have and turn it into XYZ.
In the case of sorting names, for example. As you know, to handle names well, you'll have a good bit of testing and comparison and such to arrange the name(s). and there are quite a few special cases to worry about. In this case, I'd think to myself. "Oh, this would be *so* much easier to sort if the names were split into first, middle and last name with the extra bits at the end. Then I could just sort them normally." Then I figure out how to do that.
Now, having said all that, I mainly mentioned it to help you shift your perspective on sorting. I don't know that you'll necessarily be able to use it, and the case I created above is my best guess of how it *might* be useful to you.
When your only tool is a hammer, all problems look like your thumb.